Answers are grounded in this paper and cite their sources.
Forensic paint examinations commonly rely on the interpretation of analytical data generated by multiple complementary techniques. Before implementing data fusion strategies, it is essential to understand how each analytical technique characterises the relationships between samples and whether these relationships are consistent with expert interpretation. This study investigates the similarity structure of a tool paint dataset using Fourier transform infrared spectroscopy (FTIR), Raman spectroscopy, and microspectrophotometry (MSP). Expert spectral interpretation was compared with unsupervised analyses based on principal component analysis (PCA) and hierarchical cluster analysis (HCA). Particular attention was given to the identification of indistinguishable paint groups and to the chemical factors responsible for the observed clustering patterns. The similarity relationships identified through unsupervised analyses were largely consistent with those established through expert interpretation. Most of the main paint groups identified through expert interpretation were also recovered by PCA and HCA, although some differences in sample grouping were observed. FTIR analyses revealed that the dataset was primarily structured by resin families and mineral fillers, whereas Raman analyses were more strongly influenced by pigment composition and filler materials. Based on the expert interpretation, the discriminating power obtained from each individual technique ranged from 86% to 93%, while the unsupervised analyses ranged between 86% and 89%. When analytical techniques were combined sequentially, the discriminating power increased to up to 94% for both the unsupervised analyses and expert interpretation, demonstrating the complementary nature of the information provided by FTIR, Raman spectroscopy, and MSP. These results provide a rationale for integrating the different analytical datasets through data fusion to further improve paint discrimination . Beyond the assessment of discriminating power, this study establishes a reference similarity structure for the dataset based on both expert interpretation and unsupervised analyses. These findings provide a robust foundation for evaluating the impact of future data fusion strategies and represent an essential preliminary step toward the development of multimodal approaches for forensic paint comparison.
No takes yet. Share an insight, caveat, or question.
Bedjil et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: